Guide to creating a short video

Let's create a short 5 second video based from an Image. As there are many ways you can do this, I'll show you the quickest way to get started

Creating videos is pretty easy in Mage Space, and for this article, we will go through the steps to turning a character image into a moving video. Obviously there are more ways to do it than this, but we're doing a quick short 5 second video to give you a flavor.

I'm going to use my Character Ricardo who I have previously created using the Character Builder app that is part of Mage Space. I combined various attributes and had him created as automatically. i.e. he is not a real life person.

Let's create a Character

What I'm going to do is to take this character and include him in my prompt to generate a new image. Since I'm creating a video and I also don't want him to be in that location. So in the Mage Space prompt, I type in the following prompt and also noting:

I use the @ to call on the Ricardo character which is stored.
I have to state the I want the view to be "full length", as this AI works better when you help visualise what you want to build. Otherwise, you leave it up to the AI to determine, which is usually wrong.
I could use a location if I have one stored in References, but I'm going to let AI have fun and determine this on its own.
I also make reference to his outfit as well, which again, I didn't use the references and just let Ai do its thing.

What is generated is pretty much as predicted. We have our character Ricardo jogging in a non-descript park, and he is wearing the grey sweat pants that I requested. He is also wearing the ear pods that I asked for, but I did have to made a slight edit to remove the cord.

This image will be the starting image for our video and we also have action going on in the frame, so it's a good start.

Setting up the First Frame

In order to maintain character consistency we want to create the start and the end frame. I'll show you how this is used later on, but it is essentially to make sure the video and objects flows from beginning to end and looks the way you want. I'll do the video in three setup types:

First frame only - Only have the first frame, I dictate the whole scene and let AI determine what the end looks like. This provides an unpredictable ending, but can work if you are feeling lucky.
First and last frame - I do a first and last frame to this sequence
First and alternative last frame - I do a first and last frame and give it an alternative ending to illustrate how this is treated.

For this step, it will be incredibly simple. Since we already have our first frame, we will just create a video from this. I'm going to use the Berry 2 video model for this one since I want to keep the character reference and don't want AI to warp or distort Ricardo. Without a last frame, you are pretty much leaving thing up to the machine to decide.

I include the original image in the "first frame" slot and put my prompt in, making sure to also referencing the character. Note that I don't use the "additional images" or "image" slots since these are to help the AI get a understanding of what you're after. Instead, I'm giving it a direct image for Frame 1.

Let's make "first frame only" video

As these videos will be online and playable, I'm keeping them to only 5 seconds and at 480p, a lower resolution that I would normally use.

You can see the result below which lines up with expectation. Some notes:

Mage added audio as well as sound. I didn't explicitly mention sound, but could have included "no music" in the negative prompt. It fits, so is not bad.
Video quickly does a quick camera switch to the zoomed in drinking water scene. If I wanted different, i would have asked for zooming or panning movements, but as didn't, AI made it up.

Now we will make this more coherent and make a last frame for this video so that the AI creates the space in between. I'm taking the mystery out of it by providing what I want and the result should be pretty predictive. I'm using a similar prompt with a few modifications:

Switching the video model to Melon since it does first and last frame (not all video models do this!)
On Melon, I can't use the character reference but that's okay because the first and last frame should keep consistence of character look for the few seconds.
I need to call Ricardo "man" for this prompt for the above reason, which works fine.

Let's make "first and last frame" video

The last image is this one. The background is a bit different, but he is drinking the bottle of water.

The result is what I expected, but had some changes that could not be avoided. The most jarring is the background transition which happens both because of the short time, as well as the differences between the first and last frame. But you can see that the transition is smooth and not a cut-away like the first video.

Also, I noticed that audio isn't present in this one, which is more down to the video model used and I honestly don't use the Melon model often enough. There are audio options available, but these are quite experimental at the time of typing this.

This alternate video gives Ricardo a slightly different ending. In this video, he finishes jogging and sees our other character Linda. This step is to illustrate how the video varies if a new ending is introduced, and also the wild randomness of how AI cobbles up video based on minimum prompts.

For this one, we are sticking to the first and last frame again, except:

We are back to my favourite video model, Blueberry2, which in my opinion makes the best videos, has the fewest restrictions and understands my prompts better than the others.
Because of this model, we are also restricted from using characters as well, hence the importance of the first and last frame.
I'm also spending gems here too, not much, but it's because the free minimum on this model is 4 seconds and I wanted to keep the consistency the same. Also the minimum video resolution is 540p which is why it looks slightly better.

Let's make "first and alternate last frame" video

This is the last frame we will use has a slightly different background (lighting), and also he is interacting with Linda. Ricardo is also viewed from a different side and angle as well. Let's see how the video goes.

The result shows why Blueberry 2 is the pick of the crop when it comes to video production.

Not only did it do a very smooth action transition, moving from his jogging first frame and very naturally ending up with the last frame.
Didn't seem weird or rush towards the ending. The camera panned around and changed the view to accommodate the last image and change angle.
There was no music in this, and there were human breathing sounds that sounded very natural and matched the characters.
I thought the lighting would need to be adjusted somehow as well, but the transitioned worked and didn't make it seem like it was dark and then bright all of a sudden.

For the video models that have this ability, I also ensure that the negative prompt section is populated with a standard set of criteria that you can copy from below and use.

What this feature does is that it ensures videos are created with the least amount of abnormalities and warping and the infamous "too many fingers" issues that AI generally has. it's not fool-proof, but it does help given that most video generations will need to spend gems.

Negative prompts

extra limbs, distorted hands, blurry, low quality, upside-down, upside down, flipped, inverted, rotated, sideways, tilted, handstand, headstand, melted face, broken face, distorted face, cross eyed, misaligned eyes, lopsided face, crooked jaw, smeared features, stretched skin, rubber face, extra eyes, missing eyes, wonky eyebrows, bad teeth, distorted mouth, broken neck, floating head, stretched torso, warped shoulders, misaligned body, extra limbs, missing limbs, fused arms, distorted chest, rubber skin, disproportionate, bent wrong, broken joints, deformed proportions, face warping when character moves, background music, melody, music, tunes

Guide to creating a short video

Creating videos is pretty easy in Mage Space, and for this article, we will go through the steps to turning a character image into a moving video. Obviously there are more ways to do it than this, but we're doing a quick short 5 second video to give you a flavor.

I'm going to use my Character Ricardo who I have previously created using the Character Builder app that is part of Mage Space. I combined various attributes and had him created as automatically. i.e. he is not a real life person.

Let's create a Character

What I'm going to do is to take this character and include him in my prompt to generate a new image. Since I'm creating a video and I also don't want him to be in that location. So in the Mage Space prompt, I type in the following prompt and also noting:

I use the @ to call on the Ricardo character which is stored.

I have to state the I want the view to be "full length", as this AI works better when you help visualise what you want to build. Otherwise, you leave it up to the AI to determine, which is usually wrong.

I could use a location if I have one stored in References, but I'm going to let AI have fun and determine this on its own.

I also make reference to his outfit as well, which again, I didn't use the references and just let Ai do its thing.

What is generated is pretty much as predicted. We have our character Ricardo jogging in a non-descript park, and he is wearing the grey sweat pants that I requested. He is also wearing the ear pods that I asked for, but I did have to made a slight edit to remove the cord.

This image will be the starting image for our video and we also have action going on in the frame, so it's a good start.

Setting up the First Frame

In order to maintain character consistency we want to create the start and the end frame. I'll show you how this is used later on, but it is essentially to make sure the video and objects flows from beginning to end and looks the way you want. I'll do the video in three setup types:

First frame only - Only have the first frame, I dictate the whole scene and let AI determine what the end looks like. This provides an unpredictable ending, but can work if you are feeling lucky.

First and last frame - I do a first and last frame to this sequence

First and alternative last frame - I do a first and last frame and give it an alternative ending to illustrate how this is treated.

Let's make "first frame only" video

As these videos will be online and playable, I'm keeping them to only 5 seconds and at 480p, a lower resolution that I would normally use.

You can see the result below which lines up with expectation. Some notes:

Mage added audio as well as sound. I didn't explicitly mention sound, but could have included "no music" in the negative prompt. It fits, so is not bad.

Video quickly does a quick camera switch to the zoomed in drinking water scene. If I wanted different, i would have asked for zooming or panning movements, but as didn't, AI made it up.

Now we will make this more coherent and make a last frame for this video so that the AI creates the space in between. I'm taking the mystery out of it by providing what I want and the result should be pretty predictive. I'm using a similar prompt with a few modifications:

Switching the video model to Melon since it does first and last frame (not all video models do this!)

On Melon, I can't use the character reference but that's okay because the first and last frame should keep consistence of character look for the few seconds.

I need to call Ricardo "man" for this prompt for the above reason, which works fine.

Let's make "first and last frame" video

The last image is this one. The background is a bit different, but he is drinking the bottle of water.

Let's make "first and alternate last frame" video

This is the last frame we will use has a slightly different background (lighting), and also he is interacting with Linda. Ricardo is also viewed from a different side and angle as well. Let's see how the video goes.

The result shows why Blueberry 2 is the pick of the crop when it comes to video production.

Not only did it do a very smooth action transition, moving from his jogging first frame and very naturally ending up with the last frame.

Didn't seem weird or rush towards the ending. The camera panned around and changed the view to accommodate the last image and change angle.

There was no music in this, and there were human breathing sounds that sounded very natural and matched the characters.

I thought the lighting would need to be adjusted somehow as well, but the transitioned worked and didn't make it seem like it was dark and then bright all of a sudden.

Negative prompts

Start Doing AI

Guide to creating a short video

Creating videos is pretty easy in Mage Space, and for this article, we will go through the steps to turning a character image into a moving video. Obviously there are more ways to do it than this, but we're doing a quick short 5 second video to give you a flavor.

I'm going to use my Character Ricardo who I have previously created using the Character Builder app that is part of Mage Space. I combined various attributes and had him created as automatically. i.e. he is not a real life person.

Let's create a Character

What I'm going to do is to take this character and include him in my prompt to generate a new image. Since I'm creating a video and I also don't want him to be in that location. So in the Mage Space prompt, I type in the following prompt and also noting:

I use the @ to call on the Ricardo character which is stored.

I have to state the I want the view to be "full length", as this AI works better when you help visualise what you want to build. Otherwise, you leave it up to the AI to determine, which is usually wrong.

I could use a location if I have one stored in References, but I'm going to let AI have fun and determine this on its own.

I also make reference to his outfit as well, which again, I didn't use the references and just let Ai do its thing.

What is generated is pretty much as predicted. We have our character Ricardo jogging in a non-descript park, and he is wearing the grey sweat pants that I requested. He is also wearing the ear pods that I asked for, but I did have to made a slight edit to remove the cord.

This image will be the starting image for our video and we also have action going on in the frame, so it's a good start.

Setting up the First Frame

In order to maintain character consistency we want to create the start and the end frame. I'll show you how this is used later on, but it is essentially to make sure the video and objects flows from beginning to end and looks the way you want. I'll do the video in three setup types:

First frame only - Only have the first frame, I dictate the whole scene and let AI determine what the end looks like. This provides an unpredictable ending, but can work if you are feeling lucky.

First and last frame - I do a first and last frame to this sequence

First and alternative last frame - I do a first and last frame and give it an alternative ending to illustrate how this is treated.

Let's make "first frame only" video

As these videos will be online and playable, I'm keeping them to only 5 seconds and at 480p, a lower resolution that I would normally use. You can see the result below which lines up with expectation. Some notes:

Mage added audio as well as sound. I didn't explicitly mention sound, but could have included "no music" in the negative prompt. It fits, so is not bad.

Video quickly does a quick camera switch to the zoomed in drinking water scene. If I wanted different, i would have asked for zooming or panning movements, but as didn't, AI made it up.

Now we will make this more coherent and make a last frame for this video so that the AI creates the space in between. I'm taking the mystery out of it by providing what I want and the result should be pretty predictive. I'm using a similar prompt with a few modifications:

Switching the video model to Melon since it does first and last frame (not all video models do this!)

On Melon, I can't use the character reference but that's okay because the first and last frame should keep consistence of character look for the few seconds.

I need to call Ricardo "man" for this prompt for the above reason, which works fine.

Let's make "first and last frame" video

The last image is this one. The background is a bit different, but he is drinking the bottle of water.

Let's make "first and alternate last frame" video

This is the last frame we will use has a slightly different background (lighting), and also he is interacting with Linda. Ricardo is also viewed from a different side and angle as well. Let's see how the video goes.

The result shows why Blueberry 2 is the pick of the crop when it comes to video production.

Not only did it do a very smooth action transition, moving from his jogging first frame and very naturally ending up with the last frame.

Didn't seem weird or rush towards the ending. The camera panned around and changed the view to accommodate the last image and change angle.

There was no music in this, and there were human breathing sounds that sounded very natural and matched the characters.

I thought the lighting would need to be adjusted somehow as well, but the transitioned worked and didn't make it seem like it was dark and then bright all of a sudden.

Negative prompts

Start Doing AI

As these videos will be online and playable, I'm keeping them to only 5 seconds and at 480p, a lower resolution that I would normally use.

You can see the result below which lines up with expectation. Some notes: